From: Brad Templeton (brad@main.templetons.com)
Date: Fri Nov 05 1999 - 19:19:04 CST
On Fri, Nov 05, 1999 at 12:12:37PM +0000, Charles Lindsey wrote:
> In <19991104122952.57985@main.templetons.com> Brad Templeton <brad@main.templetons.com> writes:
>
> But I am not sure what you gain from those parameters. They all have the
> same BI (they are all the same spam). What other than 'hide' do you want?
Actually global parameters would not doubt appear on the bulkcancel control
header, not on a per-message-id basis. You allow per-message-id paramters
for things that are particular to the message-id. For things that are
particular to the whole message, you put them at the top, typically.
To duplicate the form of NoCem it would be:
Control: bulkcancel; type=hide
<messageid1> ;g=groupname,groupname
<messageid2> ;g=groupname
However, this is an extensible format more in line with MIME and USENET
syntaxes and styles. We want to do that because we don't want to have
a lot of different syntaxes in the standard.
>
> The implication there is that we have not specified any usage of
> header-parameters in existing headers, notably in the Control header (for
> reasons of compatibility with existing software). You will find the syntax
The key here is that you don't allow use of extensions somewhere it might
break an old parser, but surely that doesn't apply to a new control message,
which can't possibly be being parsed by an older parser.
> Well we have no such Distribution rule yet. Does anyone else want it (I
> could write it into the Control message rules quite easily)?
I don't know what's been put in. The change is subtle but important. In
some ways, it is pointed out INN has done this to a degree. It's more a
change in style with a subtle change in the standard. A distribution
is something you are a member of. The old thing that meant was that your
feeders checked articles to see if their downstream was a member, and filtered
feeds. That didn't work. The better meaning is that sites and/or users
track what distributions they belong to, and use that to decide whether
to execute control messages, store messages and read messages.
With this semantic, a control message like cancel with a distribution on it
you do not belong to is not executed, which is the correct behaviour, I think.
>
> I would have expected corporate subnets to know their own topology without
> the use of 'sendsys' messages. The only usage I could think of would be for
Hah!